Next Article in Journal
A Novel Method and System Implementation for Precise Estimation of Single-Axis Rotational Angles
Previous Article in Journal
A Quantitative Comparison of 31P Magnetic Resonance Spectroscopy RF Coil Sensitivity and SNR between 7T and 10.5T Human MRI Scanners Using a Loop-Dipole 31P-1H Probe
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Sensors
Volume 24
Issue 17
10.3390/s24175794
sensors-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Forage Height and Above-Ground Biomass Estimation by Comparing UAV-Based Multispectral and RGB Imagery

by
Hongquan Wang
,
Keshav D. Singh
*,
Hari P. Poudel
,
Manoj Natarajan
,
Prabahar Ravichandran
and
Brandon Eisenreich
Lethbridge Research and Development Centre, Agriculture and Agri-Food Canada (AAFC), 5403 1st Avenue South, Lethbridge, AB T1J 4B1, Canada
*
Author to whom correspondence should be addressed.
Sensors 2024, 24(17), 5794; https://doi.org/10.3390/s24175794
Submission received: 21 July 2024 / Revised: 23 August 2024 / Accepted: 31 August 2024 / Published: 6 September 2024
(This article belongs to the Section Remote Sensors)
Browse Figures
Versions Notes

Abstract

:
Crop height and biomass are the two important phenotyping traits to screen forage population types at local and regional scales. This study aims to compare the performances of multispectral and RGB sensors onboard drones for quantitative retrievals of forage crop height and biomass at very high resolution. We acquired the unmanned aerial vehicle (UAV) multispectral images (MSIs) at 1.67 cm spatial resolution and visible data (RGB) at 0.31 cm resolution and measured the forage height and above-ground biomass over the alfalfa (Medicago sativa L.) breeding trials in the Canadian Prairies. (1) For height estimation, the digital surface model (DSM) and digital terrain model (DTM) were extracted from MSI and RGB data, respectively. As the resolution of the DTM is five times less than that of the DSM, we applied an aggregation algorithm to the DSM to constrain the same spatial resolution between DSM and DTM. The difference between DSM and DTM was computed as the canopy height model (CHM), which was at 8.35 cm and 1.55 cm for MSI and RGB data, respectively. (2) For biomass estimation, the normalized difference vegetation index (NDVI) from MSI data and excess green (ExG) index from RGB data were analyzed and regressed in terms of ground measurements, leading to empirical models. The results indicate better performance of MSI for above-ground biomass (AGB) retrievals at 1.67 cm resolution and better performance of RGB data for canopy height retrievals at 1.55 cm. Although the retrieved height was well correlated with the ground measurements, a significant underestimation was observed. Thus, we developed a bias correction function to match the retrieval with the ground measurements. This study provides insight into the optimal selection of sensor for specific targeted vegetation growth traits in a forage crop.

1. Introduction

Plant height and biomass are two essential biophysical parameters to quantify forage growth conditions and yield. The accurate estimation of forage height and biomass holds significant importance for plant breeders in making selection decisions, livestock management, pasture productivity assessment, and ecological monitoring. The traditional approaches for measuring crop physical parameters are destructive, time-consuming, labor-intensive, and highly subjective. With the development of smart sensors and image processing approaches, drone-based remote sensing provides a powerful tool to efficiently monitor forage growth at various temporal and spatial scales [1,2].
Compared to satellite platforms, drone-based multispectral imaging (MSI) systems are more flexible to monitor a landscape with designed imaging parameters such as flight altitude, speed, and observation angles [3]. For the data collected by low-altitude systems, the atmospheric influences on the images are also reduced, thus avoiding complex atmospheric corrections [4]. Furthermore, owing to very high spatial resolution in centimeters, UAV observation data are widely used in precision agriculture applications such as disease detection, crop growth monitoring, and yield prediction. On the other hand, the drone system is not appropriate for applications over vast areas, where satellite remote sensing is powerful. The acquisition of drone data is also expensive and time-consuming in contrast with satellite data access. For temporal analysis, it is challenging to maintain consistent solar radiation conditions across a time series of drone data.
Several approaches have been employed for forage height estimation using UAV data [5,6]. Structure from motion (SfM) photogrammetry [7] reconstructs 3D models from overlapping UAV images, allowing for precise measurement of forage height. Integration of RGB imagery with SfM facilitates detailed surface modeling, capturing variations in terrain and vegetation height. LiDAR sensors mounted on UAVs provide direct measurements of vegetation height by emitting laser pulses and analyzing the reflected signals. Coupled with MSI data, LiDAR enables accurate estimation of vegetation height even in complex terrain conditions [8]. Recently, the machine learning algorithms trained on UAV-derived MSI and RGB data have shown promise in predicting forage height [9]. Supervised learning models, such as random forest and support vector regression, utilize spectral and textural features extracted from imagery to infer forage height with high accuracy [10].
Accurate estimation of forage biomass is essential for optimizing livestock grazing, monitoring pasture productivity, and assessing ecosystem health. UAV-based MSI and RGB data offer valuable insights into vegetation structure and biochemical composition, enabling robust biomass estimation models. The spectral indices are commonly used to estimate the forage biomass [11]. The MSI allows for the calculation of vegetation indices, such as NDVI (normalized difference vegetation index) and NDRE (normalized difference red edge), which correlate with biomass density [12]. Integration of spectral indices with machine learning algorithms enhances the accuracy of biomass prediction models [13]. Additionally, integrating MSI and RGB imagery enables comprehensive characterization of vegetation properties, including biomass [14]. Fusion techniques combine spectral information from MSI sensors with spatial details from RGB imagery, enhancing the discrimination of biomass variations within a landscape [15].
MSI and RGB sensors provide valuable data for vegetation analysis. Multispectral sensors capture information across specific spectral bands, enabling the discrimination of vegetation types, health status, and biomass estimation. On the other hand, an RGB sensor offers high-resolution imagery for detailed feature extraction and analysis. This study evaluates the effectiveness of UAV-based MSI and RGB imaging systems to characterize Alfalfa (Medicago sativa L.) canopy height and above-ground biomass (AGB) proxy for efficient phenotypic traits estimation and cultivar selections [16]. The finding in this paper contributes to suggesting different sensors such as MSI and RGB for the UAV system to monitor forage biomass and height. We aim to identify the optimal sensor for different crop growth parameters.

2. Data and Methods

2.1. Study Site and Data Acquisition

The study site is located at an experimental farm of Agriculture and Agri-Food Canada, situated within Lethbridge Research and Development Center in Southern Alberta, Canada (49°42′31.4″ N, 112°45′34.7″ W).
We designed three randomized blocks, each of which had a total of 98 alfalfa plots planted (294 plots for three blocks). For each block, we set a 1 m distance between plots, with each plot containing 20 plants spaced 0.33 m apart. For each plot with an area of 6.27 m2, the biomass was harvested using a customized harvester on 8 August and 12 September 2022, respectively. The alfalfa height was measured using a standard ruler by randomly taking 3 samples over each plot on 12 October 2022.
Over the study site on 13 October 2022, we acquired the UAV-based MSI and RGB images [17] at 25 m flight altitude, using the Micasense (Seattle, MA, USA) RedEdge-P MSI sensor (6 bands) and DJI (Shenzhen, China) Zenmuse P1 camera (45MP RGB), respectively. The UAV flights were set with 85% and 80% for forward and side overlaps to enhance the quality and accuracy of the orthomosaic. Although there was a temporal gap between the UAV data acquisition and ground measurements of biomass, this study was designed to analyze the sensitivity of UAV MSI and RGB-derived features to ground measurements of vegetation parameters including canopy height and above-ground biomass. The height measurements were matched with the UAV data acquisitions. The original data were processed using the commercial photogrammetry software (Pix4D Mapper 4.8.4) to implement radiometric calibration and image stitching to obtain the orthomosaic map of the alfalfa experimental trial in Figure 1. Owing to the high spatial resolution, we could capture every single plot.

2.2. Methods

This study aims to estimate the above-ground biomass and canopy height for forage crops (Figure 2). We compared the performance between MSI and RGB sensors to retrieve the height and biomass information. Finally, optical sensors were suggested to measure canopy height and forage biomass for practical applications.

2.2.1. Canopy Height Estimation

Based on the orthomosaic, we generated a digital surface model (DSM), as seen in Figure 3a, at a fine spatial resolution of 1.67 cm/pixel from the MSI data using structure from motion algorithms [18,19]. This algorithm can estimate the 3D structure of a scenario from a series of 2D images [20]. In order to estimate canopy height, we also calculated the digital terrain model (DTM). However, the spatial resolution of the digital terrain model (DTM) in Figure 3b was degraded to 8.35 cm/pixel, five times lower than that of the DSM [21]. The degradation of spatial resolution impacts the performance of the crop height estimation. In addition, the higher vegetation conditions reduced the quality of the DTM, as fewer ground pixels are available to reconstruct surface terrain under the vegetation cover [22]. This leads to slight remaining effects of alfalfa planting structure in the DTM in Figure 3b.
In comparison with the MSI data, the RGB data were acquired at a higher spatial resolution of 0.31 cm/pixel. Similar to the processing of the MSIs, the RGB data resulted in a DSM at the original resolution of 0.31 cm/pixel, as seen in Figure 3c, and DTM at 1.55 cm/pixel in Figure 3d.
According to the methodology in [23], the canopy height model (CHM) can be simply estimated by calculating the difference between DSM and DTM layers as follows:
C H M = D S M − D T M
To ensure the same dimension of the right two terms in Equation (1), we aggregated the DSM from the original resolution to that of DTM [24]. This allows us to implement the subtraction procedure between the DSM and DTM, resulting in the CHM at 8.35 cm/pixel and 1.55 cm/pixel from the MSI and RGB images, respectively.
Apart from the difference in spatial resolution, the DSM/DTM from the RGB data is about 2 m lower than those from the MSI data [25]. This result may be due to the performance of the GPS units in the sensors [26]. It also suggests the necessity of real-time kinematic (RTK) positioning and ground control points (GCPs) to improve the coordinate accuracy in the altitude dimension [27].

2.2.2. Biomass Estimation

To develop the predictive model for the biomass, we calculated the NDVI and excess green (ExG) indices from the MSI and RGB data [28,29].
E x G = 2 G − B − R
where G, B, and R correspond to the normalized relative reflectance in the RGB data. Different vegetation indices were compared in [30,31] to retrieve vegetation biomass from multispectral data, and the NDVI obtained be best performance. Other vegetation indices such as red-edge vegetation index and canopy chlorophyll content index can be useful when the NDVI is saturated in dense vegetation conditions [32]. However, for the alfalfa crops, the NDVI was not saturated due to relatively lower height and biomass. Furthermore, high collinearity was found among different indices using the MSI and RGB data [33]. As the objective of this study is to evaluate different sensors for the UAV system to monitor alfalfa height and biomass, we selected the most representative NDVI and ExG indices for the MSI and RGB data analysis, respectively. These two indices were analyzed in terms of ground-measured biomass. Then, we established statistical relationships between the vegetation indices and ground measurements recorded on 8 August 2022. These statistical models were then used to retrieve the biomass on 12 September 2022.
As shown in Figure 4a, the NDVI values were computed from MSI to characterize the plant growth dynamic and to characterize the biomass. Aklilu Tesfaye and Gessesse Awoke [34] reported the saturation issue of NDVI, but for the forage crop, the small fall height and biomass range may not reach the saturation threshold [35]. To compare the performances between MSI and RGB data to estimate the biomass, we computed the ExG in Figure 4b from the RGB data.
We recognized that NDVI and ExG are traditional vegetation indices to characterize crop growth dynamics. The objective of this paper is to suggest an optimal selection of a suitable sensor between MSI and RGB to characterize different crop growth parameters such as height, biomass yield, vegetation cover, and leaf area index (LAI). We used statistical metrics such as correlation coefficient (R), root mean square error (RMSE), and mean absolute error (MAE) to evaluate the performances of the proposed approaches.

3. Results and Discussions

3.1. Forage Canopy Height from MSI and RGB Data

The CHM from both the MSI and RGB data in Figure 5 captured the spatial variability among different plots and within a single plot. This highlights the advantages of the UAV-based sensing technology to characterize plant growth at individual plots or even a single plant. The CHM from the MSI in Figure 5a is about 0.1 m higher than that from the RGB in Figure 5b. For a single plot, there are more valid retrievable pixels from RGB-derived CHM than from the MSI-derived CHM. More retrieved CHM information from RGB data suggests a potentially better result compared to the MSI.
The pixels of the canopy height model (CHM) results were averaged for each plot. The resulting plot-averaged height was compared with the ground measurements. In this study, by considering the range of ground measurements, we set two threshold criteria for the invalidity of the retrieved height component: height > maximum (60 cm) or height < minimum (5 cm, 10 cm). We consider the estimated values may contain more uncertainty if the height estimates are less than the minimum threshold.
Figure 6 shows the height estimate from MSI with the same maximum threshold but different minimum thresholds. The threshold of 5 cm obtained a higher correlation than that of 10 cm. However, in both cases, underestimations were observed. Nevertheless, as the minimum threshold changed from 5 cm to 10 cm, the underestimations were mitigated. Compared to the 10 cm minimum threshold, we obtained more pixels within a single plot when the minimum threshold was set to 5 cm. In the next parts of this paper, we consider developing empirical equations to correct the bias in height estimation.
In Figure 6, the deviations between the retrieved and measured height may be partially due to the difference in the observation locations and measurement timing. The manual height measurements were based on the average scaling of the three samples for each plot, but the height retrieval using CHM used all available pixels covered by the corresponding plot. Thus, CHM represents pixel-based height at different locations over a plot. In this experiment, we did not acquire the pre-seeding baseline DEM on bare soil. Thus, the DTM data were generated from the spatial interpolation of soil pixels in the DSM, which led to the reduced spatial resolution of the DTM. For the very high-resolution CHM, we may need to acquire baseline RGB or MSI data over the pre-seeded soil conditions.
In comparison, Figure 7a shows the estimated canopy height from the RGB data. In agreement with Figure 4, the estimates from the RGB were significantly lower than those from the MSI. In either case, the canopy height was underestimated. In order to correct the bias of the height (H) estimation, we developed a linear correction function as follows:
H = ( C H M − 0.11 ) / 0.16
By applying Equation (3) to the DSM estimate using RGB data, we obtained the final estimation of the height in Figure 7b. The bias-corrected height estimate agreed with the ground measurement (R = 0.65). However, the coefficients in Equation (3) need to be calibrated when applied to different scenarios.

3.2. Forage Above-Ground Biomass from MSI and RGB Data

As reported in [6,36], the NDVI is promising to monitor biomass changes over the crop growing period. In this study, we evaluated the NDVI results from the UAV MSI platform with a very high spatial resolution (1.67 cm). For each plot, the NDVI pixels covered were averaged and analyzed in terms of ground-measured fresh biomass (kg) per plot. The fresh biomass was measured two times on different periods and compared with the one-time point UAV image acquisition.
As shown in Figure 8, the average NDVI values present strong sensitivity to the ground-measured fresh biomass on 8 August and 12 September, respectively. This allowed us to develop statistical modeling algorithms to accurately retrieve above-ground biomass from the UAV MSI images.
Similarly, we calculated the average values of the RGB ExG index and analyzed them in terms of ground measurements in Figure 9. Compared to the MSI, the sensitivity of ExG to ground measurements was reduced. Thus, the MSI data seem more robust for biomass estimation than the RGB data.
Based on the sensitivity analysis, the NDVI was selected to retrieve the biomass, while the ExG was not considered in the following. We used the biomass ground measurements on 8 August to develop the empirical relationships with the NDVI:
B = 79.28 × N D V I 2 − 74.15 × N D V I + 16.33
The developed models between NDVI and biomass (B) were evaluated over the ground samples measurement on 12 September 2022. Figure 10a shows the comparison between the retrieved and ground-measured biomass, indicating a reasonable agreement (R = 0.71). Due to the limited number of UAV data acquisition, this result in Figure 10a was not an independent assessment of the model performance, but it suggests insights into the feasibility of the MSI imaging approach to estimate the forage biomass. The coefficients in Equation (4) need to be calibrated when applied to different sites.
Similar to the height estimation, we developed a bias correction function using data points presented in Figure 10a. After applying the bias correction, the ultimate estimated above-ground biomass (AGB) agrees with the ground measurement in Figure 10b.
A G B = ( B − 2.98 ) / 0.75
It should be mentioned that the coefficients in Equation (5) need to be recalibrated when used to different datasets over other experimental sites. This is the major limitation of the empirical models, compared to the physical models.
In this study, the biomass was measured per plot basis. Thus, the consistency in plot size may impact the statistical metrics. In further studies, we will consider standard biomass measurements over unit square meters.

4. Conclusions

This study proposes to monitor the forage crop (alfalfa) canopy height and above-ground biomass by comparing the performances between UAV-based MSI and RGB sensors.
The crop height was extracted from very high spatial resolution MSI and RGB images. Reasonable correlations were found between digitally derived canopy height and ground measurements (R = 0.50~0.65). When the height estimates were small, the uncertainty was found to be larger. Thus, it was necessary to set certain minimum thresholds in order to discard the invalid pixels. The RGB sensors obtained better retrieval performance for canopy height than the MSI sensor. However, the canopy height was significantly underestimated. To overcome this limitation, we developed bias-adjusted empirical equations in the canopy height estimation.
On the other hand, the traditional NDVI was found promising to characterize the fresh biomass yield (R = 0.73). This may be due to the relatively small size of the forage plant, which did not reach the saturation threshold found in the NDVI parameter. Compared to the ExG index from the RGB data, the NDVI from the MSI data had better performance in retrieving the above-ground biomass. This indicates the importance of the near-infrared wavelength to characterize the vegetation growth dynamics.
In this paper, we developed methods for physical parameter estimation using different UAV sensors. The UAV-based MSI and RGB imagery provided solutions to enhance precision agriculture and plant phenotyping. This study contributes to the selection of different sensors by considering the purpose of interesting crop-growing parameters such as height and biomass yield. The limitation of this study is the temporal gap between the UAV data acquisition and fresh biomass measurements. In perspective, we will acquire better UAV data and ground measurements with minimum temporal gaps to validate the statistical models in this paper. Furthermore, we plan to further investigate the potential of hyperspectral and LiDAR sensors to characterize the growth of different crop types.

Author Contributions

Conceptualization H.W., K.D.S. and H.P.P.; methodology, H.W.; software, K.D.S., M.N. and P.R.; validation, H.W., K.D.S. and B.E.; formal analysis, H.W.; investigation, H.W.; resources, K.D.S. and H.P.P.; writing—original draft preparation, H.W. and K.D.S.; writing—review and editing, H.W., K.D.S. and H.P.P.; visualization, H.W.; project administration, K.D.S. and H.P.P.; funding acquisition, H.P.P. All authors have read and agreed to the published version of the manuscript.

Funding

This research work was supported in part by the Beef Cattle Research Council (BCRC)-grant No. FRG.02.17, and Agriculture and Agri-Food Canada (AAFC), Canada.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data are contained within the article.

Acknowledgments

The authors acknowledge technical support from the AAFC forage field crew, especially Brandon Eisenreich, for field experiments and ground data collection.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Yoosefzadeh-Najafabadi, M.; Singh, K.; Pourreza, A.; Sandhu, K.; Adak, A.; Murray, S.; Eskandari, M.; Rajcan, I. Remote and Proximal Sensing: How Far Has It Come to Help Plant Breeders? Adv. Agron. 2023, 181, 279–315. [Google Scholar]
  2. Zhang, Y.; Xia, C.; Zhang, X.; Cheng, X.; Feng, G.; Wang, Y.; Gao, Q. Estimating the maize biomass by crop height and narrowband vegetation indices derived from UAV-based hyperspectral images. Ecol. Indic. 2021, 129, 107985. [Google Scholar] [CrossRef]
  3. Zhang, Z.; Zhu, L. A Review on Unmanned Aerial Vehicle Remote Sensing: Platforms, Sensors, Data Processing Methods, and Applications. Drones 2023, 7, 398. [Google Scholar] [CrossRef]
  4. Schläpfer, D.; Popp, C.; Richter, R. Drone data Atmospheric Correction Concept for Multi- and Hyperspectral Imagery—The Droacor Model. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2020, XLIII-B3-2020, 473–478. [Google Scholar] [CrossRef]
  5. Lussem, U.; Schellberg, J.; Bareth, G. Monitoring Forage Mass with Low-Cost UAV Data: Case Study at the Rengen Grassland Experiment. PFG-J. Photogramm. Remote Sens. Geoinf. Sci. 2020, 88, 407–422. [Google Scholar] [CrossRef]
  6. Wang, H.; Singh, K.; Poudel, H.; Ravichandran, P.; Natarajan, M.; Eisenreich, B. Estimation of Crop Height and Digital Biomass from UAV-Based Multispectral Imagery. In Proceedings of the 2023 13th Workshop on Hyperspectral Imaging and Signal Processing: Evolution in Remote Sensing (WHISPERS), Athens, Greece, 31 October–2 November 2023; pp. 1–4. [Google Scholar]
  7. Eltner, A.; Sofia, G. Structure from motion photogrammetric technique. In Developments in Earth Surface Processes; Tarolli, P., Mudd, S.M., Eds.; Elsevier: Amsterdam, The Netherlands, 2020; Volume 23, pp. 1–24. [Google Scholar]
  8. Lang, N.; Jetz, W.; Schindler, K.; Wegner, J.D. A high-resolution canopy height model of the Earth. Nat. Ecol. Evol. 2023, 7, 1778–1789. [Google Scholar] [CrossRef]
  9. Ren, H.; Zhao, Y.; Xiao, W.; Zhang, L. Unmanned Aerial Vehicle (UAV)-Based Vegetation Restoration Monitoring in Coal Waste Dumps after Reclamation. Remote Sens. 2024, 16, 881. [Google Scholar] [CrossRef]
  10. Li, W.; Niu, Z.; Shang, R.; Qin, Y.; Wang, L.; Chen, H. High-resolution mapping of forest canopy height using machine learning by coupling ICESat-2 LiDAR with Sentinel-1, Sentinel-2 and Landsat-8 data. Int. J. Appl. Earth Obs. Geoinf. 2020, 92, 102163. [Google Scholar] [CrossRef]
  11. Fan, Y.; Feng, H.; Yue, J.; Liu, Y.; Jin, X.; Xu, X.; Song, X.; Ma, Y.; Yang, G. Comparison of Different Dimensional Spectral Indices for Estimating Nitrogen Content of Potato Plants over Multiple Growth Periods. Remote Sens. 2023, 15, 602. [Google Scholar] [CrossRef]
  12. Kang, Y.; Hu, X.; Meng, Q.; Zou, Y.; Zhang, L.; Liu, M.; Zhao, M. Land Cover and Crop Classification Based on Red Edge Indices Features of GF-6 WFV Time Series Data. Remote Sens. 2021, 13, 4522. [Google Scholar] [CrossRef]
  13. Meshram, V.; Patil, K.; Meshram, V.; Hanchate, D.; Ramkteke, S.D. Machine learning in agriculture domain: A state-of-art survey. Artif. Intell. Life Sci. 2021, 1, 100010. [Google Scholar] [CrossRef]
  14. Wolff, F.; Kolari, T.H.M.; Villoslada, M.; Tahvanainen, T.; Korpelainen, P.; Zamboni, P.A.P.; Kumpula, T. RGB vs. Multispectral imagery: Mapping aapa mire plant communities with UAVs. Ecol. Indic. 2023, 148, 110140. [Google Scholar] [CrossRef]
  15. Fereidouni, F.; Griffin, C.; Todd, A.; Levenson, R. Multispectral analysis tools can increase utility of RGB color images in histology. J. Opt. 2018, 20, 044007. [Google Scholar] [CrossRef] [PubMed]
  16. Tedesco, D.; Nieto, L.; Hernández, C.; Rybecky, J.F.; Min, D.; Sharda, A.; Hamilton, K.J.; Ciampitti, I.A. Remote Sensing on Alfalfa as an Approach to Optimize Production Outcomes: A Review of Evidence and Directions for Future Assessments. Remote Sens. 2022, 14, 4940. [Google Scholar] [CrossRef]
  17. Cillero Castro, C.; Domínguez Gómez, J.A.; Delgado Martín, J.; Hinojo Sánchez, B.A.; Cereijo Arango, J.L.; Cheda Tuya, F.A.; Díaz-Varela, R. An UAV and Satellite Multispectral Data Approach to Monitor Water Quality in Small Reservoirs. Remote Sens. 2020, 12, 1514. [Google Scholar] [CrossRef]
  18. Özyeşil, O.; Voroninski, V.; Basri, R.; Singer, A. A survey of structure from motion. Acta Numer. 2017, 26, 305–364. [Google Scholar] [CrossRef]
  19. Maier, K.; Nascetti, A.; van Pelt, W.; Rosqvist, G. Direct photogrammetry with multispectral imagery for UAV-based snow depth estimation. ISPRS J. Photogramm. Remote Sens. 2022, 186, 1–18. [Google Scholar] [CrossRef]
  20. Zhou, X.; Xing, M.; He, B.; Wang, J.; Song, Y.; Shang, J.; Liao, C.; Xu, M.; Ni, X. A Ground Point Fitting Method for Winter Wheat Height Estimation Using UAV-Based SfM Point Cloud Data. Drones 2023, 7, 406. [Google Scholar] [CrossRef]
  21. Ali, M.E.N.O.; Taha, L.G.E.L.D.; Mohamed, M.H.A.; Mandouh, A.A. Generation of digital terrain model from multispectral LiDar using different ground filtering techniques. Egypt. J. Remote Sens. Space Sci. 2021, 24, 181–189. [Google Scholar] [CrossRef]
  22. Colovic, M.; Stellacci, A.M.; Mzid, N.; Di Venosa, M.; Todorovic, M.; Cantore, V.; Albrizio, R. Comparative Performance of Aerial RGB vs. Ground Hyperspectral Indices for Evaluating Water and Nitrogen Status in Sweet Maize. Agronomy 2024, 14, 562. [Google Scholar] [CrossRef]
  23. Douss, R.; Farah, I.R. Extraction of individual trees based on Canopy Height Model to monitor the state of the forest. Trees For. People 2022, 8, 100257. [Google Scholar] [CrossRef]
  24. Du, X.; Wu, H. Feature-aware aggregation network for remote sensing image cloud detection. Int. J. Remote Sens. 2023, 44, 1872–1899. [Google Scholar] [CrossRef]
  25. Ashapure, A.; Jung, J.; Chang, A.; Oh, S.; Maeda, M.; Landivar, J. A Comparative Study of RGB and Multispectral Sensor-Based Cotton Canopy Cover Modelling Using Multi-Temporal UAS Data. Remote Sens. 2019, 11, 2757. [Google Scholar] [CrossRef]
  26. Noviello, C.; Esposito, G.; Fasano, G.; Renga, A.; Soldovieri, F.; Catapano, I. Small-UAV Radar Imaging System Performance with GPS and CDGPS Based Motion Compensation. Remote Sens. 2020, 12, 3463. [Google Scholar] [CrossRef]
  27. Nakata, Y.; Iwasaki, K.; Shimoda, S.; Torita, H. Understanding microtopography changes in agricultural landscapes through precision assessments of digital surface models by the UAV-RTK-PPK method without ground control points. Smart Agric. Technol. 2023, 5, 100269. [Google Scholar] [CrossRef]
  28. Lu, N.; Zhou, J.; Han, Z.; Li, D.; Cao, Q.; Yao, X.; Tian, Y.; Zhu, Y.; Cao, W.; Cheng, T. Improved estimation of aboveground biomass in wheat from RGB imagery and point cloud data acquired with a low-cost unmanned aerial vehicle system. Plant Methods 2019, 15, 17. [Google Scholar] [CrossRef]
  29. Radočaj, D.; Šiljeg, A.; Marinović, R.; Jurišić, M. State of Major Vegetation Indices in Precision Agriculture Studies Indexed in Web of Science: A Review. Agriculture 2023, 13, 707. [Google Scholar] [CrossRef]
  30. Naito, H.; Ogawa, S.; Valencia, M.O.; Mohri, H.; Urano, Y.; Hosoi, F.; Shimizu, Y.; Chavez, A.L.; Ishitani, M.; Selvaraj, M.G.; et al. Estimating rice yield related traits and quantitative trait loci analysis under different nitrogen treatments using a simple tower-based field phenotyping system with modified single-lens reflex cameras. ISPRS J. Photogramm. Remote Sens. 2017, 125, 50–62. [Google Scholar] [CrossRef]
  31. Kross, A.; McNairn, H.; Lapen, D.; Sunohara, M.; Champagne, C. Assessment of RapidEye vegetation indices for estimation of leaf area index and biomass in corn and soybean crops. Int. J. Appl. Earth Obs. Geoinf. 2015, 34, 235–248. [Google Scholar] [CrossRef]
  32. Liu, Y.; Liu, S.; Li, J.; Guo, X.; Wang, S.; Lu, J. Estimating biomass of winter oilseed rape using vegetation indices and texture metrics derived from UAV multispectral images. Comput. Electron. Agric. 2019, 166, 105026. [Google Scholar] [CrossRef]
  33. Yuan, Y.; Wang, X.; Shi, M.; Wang, P. Performance comparison of RGB and multispectral vegetation indices based on machine learning for estimating Hopea hainanensis SPAD values under different shade conditions. Front. Plant Sci. 2022, 13, 928953. [Google Scholar] [CrossRef] [PubMed]
  34. Aklilu Tesfaye, A.; Gessesse Awoke, B. Evaluation of the saturation property of vegetation indices derived from sentinel-2 in mixed crop-forest ecosystem. Spat. Inf. Res. 2021, 29, 109–121. [Google Scholar] [CrossRef]
  35. Gu, Y.; Wylie, B.K.; Howard, D.M.; Phuyal, K.P.; Ji, L. NDVI saturation adjustment: A new approach for improving cropland performance estimates in the Greater Platte River Basin, USA. Ecol. Indic. 2013, 30, 1–6. [Google Scholar] [CrossRef]
  36. Cabrera-Bosquet, L.; Molero, G.; Stellacci, A.M.; Bort, J.; Nogués, S.; Araus, J.L. NDVI as a Potential Tool for Predicting Biomass, Plant Nitrogen Content and Growth in Wheat Genotypes Subjected to Different Water and Nitrogen Conditions. Cereal Res. Commun. 2011, 39, 147–159. [Google Scholar] [CrossRef]
Sensors 24 05794 g001
Figure 1. Study site with UAV multispectral imagery on 13 October 2022.
Figure 1. Study site with UAV multispectral imagery on 13 October 2022.
Sensors 24 05794 g001
Sensors 24 05794 g002
Figure 2. UAV multispectral and RGB imaging analysis workflow to estimate canopy height and above-ground biomass.
Figure 2. UAV multispectral and RGB imaging analysis workflow to estimate canopy height and above-ground biomass.
Sensors 24 05794 g002
Sensors 24 05794 g003
Figure 3. (a) DSM and (b) DTM from multispectral imagery, in comparison with (c) DSM and (d) DTM from RGB imagery.
Figure 3. (a) DSM and (b) DTM from multispectral imagery, in comparison with (c) DSM and (d) DTM from RGB imagery.
Sensors 24 05794 g003
Sensors 24 05794 g004aSensors 24 05794 g004b
Figure 4. Spatial distribution of (a) MSI NDVI and (b) RGB ExG on 13 October 2022.
Figure 4. Spatial distribution of (a) MSI NDVI and (b) RGB ExG on 13 October 2022.
Sensors 24 05794 g004aSensors 24 05794 g004b
Sensors 24 05794 g005
Figure 5. Canopy height model (CHM) from (a) MSI and (b) RGB imagery.
Figure 5. Canopy height model (CHM) from (a) MSI and (b) RGB imagery.
Sensors 24 05794 g005
Sensors 24 05794 g006
Figure 6. Validations of the retrieved height from MSI by setting the minimum threshold as (a) 5 cm and (b) 10 cm, in terms of ground measurements.
Figure 6. Validations of the retrieved height from MSI by setting the minimum threshold as (a) 5 cm and (b) 10 cm, in terms of ground measurements.
Sensors 24 05794 g006
Sensors 24 05794 g007
Figure 7. Validations of the retrieved height from RGB data: (a) original retrieval; (b) bias correction for height estimation.
Figure 7. Validations of the retrieved height from RGB data: (a) original retrieval; (b) bias correction for height estimation.
Sensors 24 05794 g007
Sensors 24 05794 g008
Figure 8. Relationships between UAV-based mean NDVI values for each plot on 13 October and measured fresh biomass on (a) 8 August and (b) 12 September 2022. Each plot was designed with an area of 6.27 m2.
Figure 8. Relationships between UAV-based mean NDVI values for each plot on 13 October and measured fresh biomass on (a) 8 August and (b) 12 September 2022. Each plot was designed with an area of 6.27 m2.
Sensors 24 05794 g008
Sensors 24 05794 g009
Figure 9. Relationships between UAV-based mean ExG values for each plot (black dots) on 13 October and measured fresh biomass on (a) 8 August and (b) 12 September 2022.
Figure 9. Relationships between UAV-based mean ExG values for each plot (black dots) on 13 October and measured fresh biomass on (a) 8 August and (b) 12 September 2022.
Sensors 24 05794 g009
Sensors 24 05794 g010
Figure 10. UAV multispectral-retrieved (a) biomass and (b) bias-corrected biomass on 12 September 2022.
Figure 10. UAV multispectral-retrieved (a) biomass and (b) bias-corrected biomass on 12 September 2022.
Sensors 24 05794 g010
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Wang, H.; Singh, K.D.; Poudel, H.P.; Natarajan, M.; Ravichandran, P.; Eisenreich, B. Forage Height and Above-Ground Biomass Estimation by Comparing UAV-Based Multispectral and RGB Imagery. Sensors 2024, 24, 5794. https://doi.org/10.3390/s24175794

AMA Style

Wang H, Singh KD, Poudel HP, Natarajan M, Ravichandran P, Eisenreich B. Forage Height and Above-Ground Biomass Estimation by Comparing UAV-Based Multispectral and RGB Imagery. Sensors. 2024; 24(17):5794. https://doi.org/10.3390/s24175794

Chicago/Turabian Style

Wang, Hongquan, Keshav D. Singh, Hari P. Poudel, Manoj Natarajan, Prabahar Ravichandran, and Brandon Eisenreich. 2024. "Forage Height and Above-Ground Biomass Estimation by Comparing UAV-Based Multispectral and RGB Imagery" Sensors 24, no. 17: 5794. https://doi.org/10.3390/s24175794

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop